Long Video Generative Performance on MovieChat-1K breakpoint mode 1.0
2.64CIMovieChat
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MovieChatEvaluation Mode=Breakpoint, Evaluation Protocol=Average of GPT-3.5, Claude, and human blind rating2023.07 | 2.64 | 2.6 | 2.87 | 2.49 | 3.08 | |
| Video-ChatGPTEvaluation Mode=Breakpoint, Evaluation Protocol=Average of GPT-3.5, Claude, and human blind rating2023.07 | 2.62 | 2.65 | 2.86 | 2.32 | 2.96 | |
| Video ChatEvaluation Mode=Breakpoint, Evaluation Protocol=Average of GPT-3.5, Claude, and human blind rating2023.07 | 2.42 | 2.51 | 2.81 | 2.1 | 2.78 | |
| Video LLaMAEvaluation Mode=Breakpoint, Evaluation Protocol=Average of GPT-3.5, Claude, and human blind rating2023.07 | 2.04 | 2.29 | 2.63 | 2 | 2.87 |